FUNA Y FLORA

Participants and Procedure

. Six-hundred men (13% Asian; 7% Black or African-American; 72% Caucasian; 7% Hispanic or Latino) ages 18–50 (Mage = 27.2; SD = 6.8) were recruited via Amazon’s Mechanical Turk (MTurk) web site. Table 1 provides demographic information. This site permits the online collection of data and typically offers greater sample diversity than typical convenience samples (Buhrmester, Kwang, & Gosling, 2011). The sample was restricted to men from the U.S. only because gender socialization is culturally driven and may differ by country. The U.S. Individuals who participated were compensated $2.00 each for completion of the questionnaires. This data set was previously utilized. Therefore, there was no need for IRB approval and consent statements for the current study. The study population of interest was adult males in the United States.

Demographic Questionnaire

Participants responded to a series of questions about age, ethnicity, marital status, relationship history, self-identified sexual orientation, and level of education.

Measures

Predictor variables

Gender role discrepancy and discrepancy stress

GRD and DS/MDS (Reidy et al., 2014) scale, respondents answered 5 Likert-type questions (1 “Strongly Agree” to 7 “Strongly Disagree”) pertaining to the experience of

perceived gender role discrepancy (i.e., “I am less masculine than the average guy,” “compared to my guy friends I am not very masculine,” “most women I know would say that I’m not as masculine as my friends,” “most guys would say I’m not very masculine compared to them,”

Page 20 of 62

“most women would consider me to be less masculine than the typical guy”) and 5 Likert-type questions about discrepancy stress (i.e., "I wish I was more manly," "I wish I were interested in things that other guys find interesting," "I worry that people judge me because I'm not like the typical man," "sometimes I worry about my masculinity," "I worry that women find me less attractive because I'm not as macho as other guys").

Abbreviated Masculine Gender Role Stress Scale

The abbreviated MGRS (Swartout, Parrott, & Cohn, 2015), utilizes the 15 items with the highest item-to-total scale correlations. Psychometric properties of each of the 15 items were examined with item response theory (IRT) analysis (Swartout et al., 2015). Item response theory results showed that the abbreviated scale might hold promise at capturing the same amount of information as the full 40-item scale (Swartout et al., 2015). The abbreviated MGRS scale assessed men’s experience of stress associated with events related to the male gender role, including physical inadequacy, expressing tender emotions, subordination to women, intellectual inferiority, and failure at work or sex (Eisler & Skidmore, 1987). Respondents rate situations according to how stressful they feel, if involved in each situation. Scoring is done by using a 6-point Likert scale (0 “not at all stressful” to 5 “extremely stressful”), with higher scores indicating greater masculine role stress, (i.e., “Being outperformed at work by a woman” (Subordination to Women), “Being perceived by someone as gay” (Physical Inadequacy), “Admitting that you are afraid of something” (Emotional Inexpressiveness), “Having people say that you are indecisive” (Intellectual Inferiority), “ Getting passed over for a promotion” (Performance Failure).

Page 21 of 62

Male Role Norms Sub-Scale

In examining MRNS (MRNS; Thompson & Pleck, 1986), according to Moore and Stuart (2005), the MRNS adheres to the normative approach of measuring masculinity in that “rather than examining how men describe themselves, this approach examines masculinity in terms of men’s beliefs about how men and women should think, feel, and behave, as well as their rights and roles in society”. Scores on the MRNS correlate with men and women’s attitudes toward men and are inversely related to attitudes of gender equality (Thompson & Pleck, 1995). This scale is a 26-item Likert-type scale (1 "strongly disagree” to 7 “strongly agree") that measures traditional Western masculine ideology relating to status, toughness, and antifemininity.

Outcome Variables

Psychiatric History Each participant’s psychiatric history was measured via one indicator. Participants

indicated (a) have you been diagnosed with a mental illness? Participants answered using 1 for “yes” and 2 for “no.”

Pathological Gambling

Participants indicated how many times they had "lost more money than you can afford." Participants answered using a discrete (count) response.

Substance Use

Each participant’s substance use disorder was measured via six indicators taken from the Crime and Analogous Behavior scales (CAB; Miller & Lynam, 2003). Participants indicated how many times per month they had (a)"consumed alcohol, until intoxicated” and (b) “used illegal

Page 22 of 62

drugs, (marijuana, hashish, cocaine, ecstasy, mushrooms, heroin, speed, pills, etc.).” Participants answered how old when you first (c) “got “buzzed” or drunk” and (d)” used illegal drugs.” Participants answered how many times you have (e) driven while “buzzed”, drunk or high and (f) “How many times have been arrested for driving under the influence of drugs or alcohol in your lifetime.” Participants answered using discrete (count) responses for items; a, b, e, and f and answered using continuous responses for; c and d.

Violence

Each participant’s violence was measured via four indicators taken from the CAB scale (Miller & Lynam, 2003). Participants indicated how many times have you (a) "been in a physical fight with another", (b) "attacked someone with the intent to cause harm, injure, rape, or kill," (c) "attacked someone with a weapon with intent to harm, injure, rape, or kill" and (d) “intentionally hurt someone to a degree that he/she needed bandages or a doctor.” Participants answered using discrete (count) responses.

Sexual Risk Behavior

Each participant’s sexual risk behavior was measured via four indicators taken from the CAB scale (Miller & Lynam, 2003). Participants indicated how old were you when you first (a) “intercourse (vaginal or anal)”. Participants indicated how many (b)” partners have you had sexual intercourse with,” (c) “times have you had unprotected intercourse, with someone you were not in a relationship with” and (d) “times have you been diagnosed with a sexually transmitted disease.” Participants answered using discrete (count) responses on items; b, c, and d and used continuous responses for item a.

Page 23 of 62

Data Analysis

All analyses employed structural equation modeling (SEM) in Mplus version 8.0.

Structural equation modeling takes factor analysis one step further by relating the constructs to each other and to covariates in a system of linear regressions thereby purging the biasing effects of measurement error. All robust estimation and missing data were handled using maximum likelihood parameter estimates (MLR) for nonbinary data. Maximum likelihood parameter estimates were utilized because this estimator leads to better standard errors, due to utilization of weights.

Phase One: Assessing Measurement Models

During the first phase measurement models were tested through confirmatory factor analysis. This is done to describe the relationship between observed variables (e.g.,

instruments) and the construct or constructs those variables are hypothesized to measure and ultimately determine a best fitting model. Models were deemed to fit the underlying data adequately when the root mean square error of approximation (RMSEA) reached 0.08 and the Comparative Fit Index (CFI)/Tucker Lewis Index (TLI) reached .90. For those measurement models that could not achieve adequate fit, exploratory factor analysis (EFA) would be conducted, utilizing the chisquaredifference test, to identify the best fitting factor structure. Phase Two: Assessing the Superordinate Measurement Model

During the second phase, we tested the existence of the superordinate model. This is an aggregate or a second order latent variable composed of the outcomes, psychiatric history, pathological gambling, substance use, violence, and sexual risk behavior. See Figure 1 below for conceptual Superordinate Measurement Model that was tested.

Page 24 of 62

The presence of a superordinate model would be confirmed when adequate measure is achieved with root means square error of approximation (RMSEA) reaching 0.08 and the Comparative Fit Index (CFI)/Tucker Lewis Index (TLI) reaches .90.

Phase 3: The Full Model(s)

If an adequate superordinate measurement model is obtained, the next step is regressing the higher order psychosocial adjustment factor on our predictor constructs, demonstrating their associations to the shared variance of the five first order outcomes. See Figure 2 below for the conceptual Superordinate SEM that will be tested.

Psychosocial Adjustment Substance Use Violence Sexual Risk Behavior Psychological Diagnosis Gambling

Page 25 of 62

If the superordinate measurement model is unidentified, we will eliminate the

superordinate factor and measure the influence of four predictors GRD, MDS, MGRS, and MRNS (antifemininity) on seven independent outcomes, delinquency onset, substance use, sexual risk behavior, violence, STD, pathologic gambling, and psychiatric diagnosis. This model aimed to regress each outcome on each predictor, creating a multivariate multiple regression SEM. See Figure 3 below for the conceptual Full SEM that was tested.

GRD MDS MGRS MRNSa Psychosocial Adjustment Psychological Diagnosis Gambling Substance Use Sexual Risk Behavior Violence GRD DELIQ VIOL SRB SU MDS MRNSa MGRS STD PSYCHDX GAMB

Page 26 of 62